Snowflake Pricing Explained: A Complete Guide for 2026

Let’s face it: the cloud is complex. It’s become more complex in recent years, and it will likely be more complex in years to come. So it should come as no surprise that operating a cloud data platform like Snowflake and managing its compute consumption can be complex.

Multiple factors determine your monthly Snowflake spend. These include pricing tiers, warehouse sizes, query latency, credit discounts, and more. Some of those factors are set in your contract. Others change based on usage and performance.

If you’re going to optimize your Snowflake spend and performance, it helps to know where that spend is going. In this guide, we’ll break down exactly how Snowflake pricing works and how to choose the best setup for your organization.

Note: The most authoritative source for Snowflake pricing is their Service Consumption Table. This is the primary source we’ve used throughout this article.

How Is Snowflake Pricing Calculated?

Snowflake pricing is based on usage. This model has three components:

• Storage costs are calculated at a flat rate per terabyte (TB) per month, measured on the compressed data Snowflake stores. That includes database tables, files staged for bulk loading and unloading, and historical data kept for Time Travel and Fail-safe.
• Compute costs are incurred any time you consume Snowflake credits. They fall into three categories: virtual warehouses, serverless, and cloud services. AI features now bill separately in AI Credits.
• Data transfer costs are incurred when you transfer data from a Snowflake account to a different region, cloud provider, or both.

Each of these pricing components has its own nuances and complexities. Let’s take a closer look at each.

Note: Snowflake offers a free trial, which includes $400 worth of credits. That means you have the option to test the platform and see if it works for you before making a commitment.

Storage Costs

Snowflake stores data in compressed, columnar micro-partitions that sit in the underlying cloud provider’s storage layer (AWS, GCP, or Azure). Micro-partitions are contiguous units of storage used to organize and manage data within Snowflake tables. Each holds 50 to 500 MB of uncompressed data.

Snowflake bills storage on the average number of on-disk bytes stored each day. It takes periodic snapshots of your stored data, calculates a daily average, and bases the monthly charge on the average across that month’s calendar days. Because those bytes are compressed, you pay for far less than the raw size of the data you loaded.

The exact cost varies by region, cloud platform, and whether the account is on a Capacity or On-Demand contract. Generally speaking, expect storage to cost less in the U.S. than in Europe or Asia. Likewise, Capacity storage prices are significantly lower than On-Demand. In US regions, that’s roughly $23/TB versus $40/TB per month, which adds up quickly across a large database.

Note: Snowflake compresses data automatically, and you’re billed on the compressed size. Compression ratios depend on your data, but they’re often substantial. Snowflake’s own cost examples assume 5:1.

Data included in Snowflake storage costs include:

• Files staged for bulk loading/unloading (compressed or uncompressed)
• Database tables
• Historical data stored for Time Travel
• Fail-safe data for permanent tables
• Clones of database tables that reference data deleted from their source tables
• Materialized views and search optimization access paths

Iceberg tables are the exception. Their data lives in your own cloud storage, so your cloud provider bills for that storage rather than Snowflake.

Compute Costs

Warehouse Costs

Snowflake runs its data operations on virtual warehouses. These vary in size based on the resources provisioned to them. Thankfully, Snowflake warehouse sizes are easy to follow; as easy as T-shirt sizes. Each increase in size doubles the credits consumed per hour.

Snowpark-optimized warehouses provide much more memory per node for memory-heavy work like ML training and large Python UDFs. They bill at 1.5 times the standard rate for the same size. The 5X-Large and 6X-Large sizes aren’t available on every cloud and region, so check availability for your account before planning around them.

Some other key considerations when calculating warehouse costs:

• Warehouses consume credits only while running. A suspended warehouse costs nothing.
• Snowflake charges for a minimum of 60-seconds, regardless of actual runtime. After the first minute, Snowflake bills per second. For larger warehouses that start and stop often, those minimums can become a source of cost inefficiencies. 
• “Bigger” doesn’t automatically mean “more expensive.” For workloads that parallelize well, doubling a warehouse’s size can cut run times roughly in half, so the task costs about the same. Although, this isn’t always the case across every workload. 
• Multi-cluster warehouses multiply the rate. On Enterprise Edition and above, a warehouse can add clusters under concurrency pressure. Each running cluster bills at the full rate for its size.
• Suspension clears the cache. While a warehouse runs, it caches table data so later queries don’t have to constantly re-read from remote storage. When the warehouse suspends, that cache is lost, causing subsequent queries to run slower.

We’ll get into what a credit actually costs in our section on Snowflake credits below.

Serverless Costs

Serverless features run on Snowflake-managed compute instead of your virtual warehouses. They bill separately from warehouse usage and appear as their own line items. Rather than a single serverless price, Snowflake charges for each feature based on its billing basis, your usage volume, and your credit price.

Most serverless features bill in compute-hours. One compute-hour equals the resources of an X-Small warehouse running for one hour. Snowflake charges one credit per compute-hour, then applies a multiplier specific to each feature. Some features add cloud services charges or unit charges on top of that.

Note that for serverless features:

• There’s no 60-second minimum; serverless bills per second
• Serverless cloud services do not count toward your 10% cloud services adjustment
• Some features, like Snowpipe and Optima Clustering, bill by volume (per gigabyte) rather than per compute-hour

What a serverless credit costs in dollars depends on your Snowflake edition, cloud provider, region, and whether you buy credits at Capacity or On-Demand (see below). For the table below, we’ll assume $3/credit, the on-demand Enterprise rate in AWS US East.

So, as an example, a team ingesting 5,000 GB of CSV files a month through Snowpipe consumes 18.5 credits, or $55.50 at $3 per credit.

If you’ve read older Snowflake pricing guides, you may have seen materialized views and search optimization listed at “10 credits per hour.” That figure comes from a prior version of the Consumption Table, and is no longer accurate.

A few other details are worth knowing:

• Snowpipe has moved to per-GB pricing. It used to combine per-second compute with a per-1,000-files fee. Business Critical and VPS accounts switched on August 1, 2025; Standard and Enterprise accounts followed on December 8, 2025.
• Maintenance costs track data churn, not query volume. Clustering, materialized views, and search optimization do their work when the underlying tables change. A high-churn table drives high maintenance charges even if no one queries it.
• Search optimization carries a storage cost. The search access path is typically about one-quarter the size of the source table.
• Dynamic tables aren’t serverless. Their refreshes run on a virtual warehouse you assign, so they show up as warehouse spend.

Cloud Services Costs

Snowflake’s cloud services layer handles all the functionality of the platform, except for the actual storing and processing of data. This includes:

• Authentication
• Infrastructure management
• Metadata management
• Query compilation and optimization
• Access control
• Zero-copy cloning

Cloud services consume credits like everything else, but Snowflake waives most of them. Each day, it applies an adjustment equal to the lesser of two numbers: your cloud services credits, or 10% of that day’s warehouse credits. You only pay for what’s left over.

Here’s how that plays out on two different days:

Two details typically trip people up. First, the 10% is measured against warehouse usage only, not your total compute, so serverless and AI spend don’t raise the threshold. Second, the adjustment is calculated day by day. One day of heavy usage still gets charged, even if the rest of the month is slow. 

In our experience, most accounts stay under the threshold; cloud services have little to no impact on users’ final bills. But there are exceptions to this rule. Environments with the following patterns run a higher risk of exceeding it:

• Frequent schema changes or DDL-heavy pipelines
• Large volumes of very short queries, where compilation outweighs execution
• Heavy role-based access control activity
• Large-scale cloning operations

So if your warehouse credit usage remains stable but your overall bill is climbing, investigate cloud services through the METERING_DAILY_HISTORY view in the ACCOUNT_USAGE schema. It shows used, adjusted, and billed cloud services credits side by side.

Data Transfer Costs

Every Snowflake account runs in one of Snowflake’s supported regions on one of the three major cloud providers. Data transfer charges apply when data leaves that region or cloud. That’s why most Snowflake customers place their primary account in the same region and cloud as the systems that feed and consume it.

Keep the following in mind when considering Snowflake data transfer costs:

• Ingress is free. Snowflake doesn’t charge to bring data into the platform.
• Same region, same cloud is free. Moving data within one region on one cloud provider incurs no transfer charge.
• Query results aren’t egress. Clients retrieving query results don’t trigger data transfer fees.
• Cross-region and cross-cloud egress is billed per TB. The rate depends on the source region, the destination, and the cloud provider.
• Only certain operations incur charges. Unloading data, replication, external functions, and a handful of others (see Snowflake’s full list).

Common examples that do incur transfer charges:

• Unloading data to an S3 bucket in a different region
• Replicating a database from AWS US East to Azure West Europe for disaster recovery
• Sharing data with a consumer account in another region or cloud

Pro tip: Align your Snowflake region with the regions of your primary downstream data consumers. Daily cross-region replication of a large database is where transfer fees stop being a rounding error.

Snowpark Container Services Costs

Snowpark Container Services (SPCS) is Snowflake’s fully managed container offering. With SPCS, you can run containerized workloads, like model serving, custom apps, and APIs, directly inside Snowflake.

Instead of virtual warehouses, SPCS runs on compute pools: groups of nodes built from a chosen instance family. A compute pool bills credits per hour for every node that’s running, whether or not your services are busy. Like warehouses, pools can auto-suspend, and that setting is the biggest lever on SPCS spend.

AI Costs: Snowflake AI Credits

This is the biggest change to Snowflake pricing in the past year. Snowflake’s AI features no longer consume the same credits as your warehouses. They bill in AI Credits, a separate unit from the platform credits that cover warehouses, serverless features, and cloud services.

AI Credits are priced differently from platform credits in three important ways:

• One price, every edition. An AI Credit costs $2.00 with global routing and $2.20 with regional routing, whether you’re on Standard or Business Critical.
• No capacity discount. The discount you negotiate on platform credits doesn’t apply to AI Credits.
• Routing keeps processing in your region. Global routing lets Snowflake process a request in whichever region has model capacity. Regional routing keeps processing in your region, at a 10% premium, which matters if you have data residency requirements.

Features that bill in AI Credits include:

• Cortex Agents, for building agents that reason over your structured and unstructured data
• AI Functions, which call large language models from SQL
• Cortex Search, for hybrid search over your data
• AI Parse Doc, which extracts content from documents and bills per 1,000 pages
• Snowflake CoWork

Each feature meters its own usage, and Snowflake converts that usage into AI Credits. Current rates per feature and per model live in Snowflake’s Cortex AI pricing documentation.

If you’re working from an older guide, note what’s changed. Per-token Cortex tables tied to specific models, and Document AI billed per second, no longer reflect how Snowflake AI pricing works.

One more thing to watch: AI Credits cover the AI processing, not everything around it. The warehouse that runs a query calling an AI function still bills platform credits as usual. A complete forecast for an AI workload needs both.

What Are Snowflake Credits and How Are They Priced?

We’ve mentioned Snowflake credits enough times now that we should stop and answer the question directly.

Snowflake credits are the unit Snowflake uses to measure and charge for compute. Platform credits are consumed by:

• Virtual warehouses
• Serverless features
• Snowpark Container Services compute pools
• Cloud services (only the portion above the daily 10% adjustment)

AI features consume AI Credits instead, priced separately at $2.00 to $2.20 (see above).

How Much Does a Snowflake Credit Cost?

On-demand platform credit prices fall in these ranges:

These are list prices, not averages. Four factors decide where you land:

• Edition. Each tier up carries a higher base rate.
• Cloud provider. AWS, Azure, and GCP are priced separately.
• Region. US regions sit at the low end of each range. Other regions, particularly in Europe and Asia-Pacific, sit higher.
• Contract. A capacity commitment lowers your rate below these on-demand prices.

To check the exact rate for your setup, use Snowflake’s pricing page and calculator.

Here’s how credits turn into dollars. Say you’re on Enterprise in AWS US East at $3.00 per credit, and a Medium warehouse runs 70 hours this month. That’s 4 credits/hour × 70 hours = 280 credits, or $840.

You can buy Snowflake credits in one of two ways: on-demand or capacity.

On-Demand

On-demand is a true pay-as-you-go model. Snowflake bills you monthly in arrears for what you consume, at the list prices above. If you’re unsure how many credits you’ll need, or confident your usage will stay small, on-demand is often the best choice.

• Pros: Your Snowflake spend is directly proportional to your usage, with no commitment. That’s helpful when you don’t yet know how much compute you need.
• Cons: Snowflake workloads are dynamic and difficult to predict. A workload you expected to be small can grow through inefficient queries, rising concurrency, or new use cases. On-Demand also carries the highest per-credit price, which adds up quickly across large workloads.

Capacity

With capacity, you commit upfront to a dollar amount of Snowflake usage over a contract term, typically a year or more. In exchange, you get a lower rate on platform credits and storage. Capacity contracts start at $25,000. Snowflake doesn’t publish its discount tiers; they’re negotiated and generally improve with the size and length of the commitment.

If you use up your commitment before the term ends, additional usage bills at on-demand rates until you renew or buy more. Snowflake says customers may be able to carry unused capacity into a future period, generally when they purchase additional capacity at renewal. That isn’t a promise of rollover on every renewal, and there’s no universal percentage, deadline, or minimum new commitment. 

• Pros: The biggest pro is a lower per-credit price. If your workloads are predictable, or you can forecast usage over the contract term, capacity is a strong way to keep costs under control.
• Cons: If your projections are too high, you end up over-provisioning credits you don’t use. If they’re too low, the overage bills at on-demand prices. Finding the right balance is a delicate exercise, and the dynamic nature of cloud workloads makes it harder. An unexpected usage spike can push you past your commitment months early, which turns a budgeting exercise into a conversation with finance.

You can see how that plays out in dollars in our pricing example below.

What Are the Snowflake Pricing Tiers?

As the table above shows, your credit price depends on which Snowflake Edition you run. Each Edition adds features on top of the one below it, and the right choice usually comes down to your security, compliance, and concurrency needs. Moving up a tier raises the price of every credit you consume, so it’s worth choosing deliberately.

Snowflake Standard Edition

Standard is the entry-level tier, and it covers the core platform: virtual warehouses, data loading, query processing, data sharing, and up to one day of Time Travel. It works well for smaller teams and straightforward analytics workloads that don’t need advanced governance or scaling features.

Snowflake Enterprise Edition

Enterprise is the most common choice for larger organizations. It adds multi-cluster warehouses for handling concurrency, up to 90 days of Time Travel, materialized views, search optimization, and more granular governance controls like column-level security. Several of those features carry their own serverless or storage costs, which is one more reason to track feature usage, not just warehouse hours.

Snowflake Business Critical Edition

Organizations with highly sensitive data, such as healthcare companies handling PHI under HIPAA, should seriously consider Business Critical. It adds stronger encryption options, private connectivity to cloud provider networks, support for regulated workloads, and database failover and failback for disaster recovery across regions.

Snowflake Virtual Private Snowflake (VPS) Edition

VPS is Snowflake’s highest tier. It runs in a completely separate Snowflake environment that shares no hardware or services with accounts outside it. It’s built for organizations in highly regulated industries, like financial services, that need the strictest isolation available. At roughly 3x the Standard credit price, it’s a tier you choose when compliance matters more than convenience. 

Snowflake Pricing Example

Now that we’ve laid out how each piece of Snowflake pricing works, let’s see an example of how they all fit together. 

Consider a company on the Enterprise plan, deployed on AWS in US East, that commits to $30,000 of capacity each month. Snowflake doesn’t publish its discounts, so let’s assume this contract lands at 20% off the $3.00 on-demand rate. That’s $2.40 per credit, or 12,500 credits a month.

This company also stores an average of 5 TB of compressed data. At the $23/TB capacity rate, that’s $115 in storage, bringing the planned spend to $30,115 per month.

Here’s the anticipated usage:

That leaves 1,130 credits available to absorb normal fluctuations in usage. On paper, the capacity plan is tightly aligned with usage.

Now consider what happens with an unanticipated influx of users. A performance SLA requires the team to scale all fifteen warehouses from Medium to X-Large to hold query latency. It happens three times a day, every other day, for about 10 minutes each time.

That’s 3 spikes × 15 days × 10 minutes = 7.5 hours a month at X-Large. Because the spikes fall within the warehouses’ normal running hours, the extra cost is the difference between the two sizes: 16 − 4 = 12 credits/hour. Across fifteen warehouses, that’s 15 × 12 × 7.5 = 1,350 additional credits.

The first 1,130 are covered by wiggle room. The remaining 220 bill at the $3.00 on-demand rate.

If the team had planned for the spikes and committed to 12,720 credits instead, the bill would be $30,528 for credits plus $115 for storage, or $30,643. That’s $132 a month, or $1,584 a year, lost to the on-demand premium on just 220 credits.

Commonly Overlooked Behaviors That Drive Up Snowflake Costs

Snowflake’s pricing model is flexible by design, which means your bill reflects how you configure and use the platform. In our experience, most budget surprises happen because users make choices that seem right at the time, but cause problems as workloads evolve. 

Oversized Warehouses

When a team isn’t sure how demanding a workload will be, it’s tempting to default to Medium or Large “to be safe.” But each size up doubles the hourly rate. If most queries on that warehouse finish in a few seconds, a smaller size would often run them nearly as fast for a fraction of the credits. Applied across thousands of queries a day, this can become a significant cost multiplier.

Auto-Suspend Settings That Work Against You

Auto-suspend is essential to controlling costs, but setting that threshold too low can backfire on you. Each time a warehouse suspends, it loses its local cache, so the next queries re-read data from remote storage. Each time it resumes, the 60-second minimum starts again. A warehouse that cycles on and off every few minutes ends up paying for each 60-second restart plus the cold-storage data. 

Serverless Features on Tables That Don’t Need Them

Search optimization, materialized views, and automatic clustering can be valuable when applied to the right tables. They don’t run on a schedule; instead, they do maintenance work whenever the underlying data changes. That means a feature turned on for one use case keeps billing as long as the table keeps churning, even after the dashboard that needed it is gone.

A quick check to see if you’re paying for unneeded serverless features:

• Is search optimization enabled on archive tables nobody searches anymore?
• Do materialized views sit on high-churn staging tables that are rarely queried?
• Is replication running to a region where no one is actively working?

Capacity Commitments That Don’t Match Usage Patterns

As the pricing example shows, a capacity plan that’s slightly off increases your costs. Underestimate, and overage bills at on-demand rates. Overestimate, and you’ve prepaid for credits you may not be able to carry forward. Neither forecast stays accurate for long when workloads change month to month.

Cross-Region and Cross-Cloud Transfers

Transfer fees look small per terabyte, so they rarely get attention when you’re making your initial architectural decisions. However, they add up significantly when replication or unloads run daily across regions or clouds. Map your pipelines to your region strategy before they’re built, and don’t assume multi-cloud capability means free data movement.

AI Usage Without Guardrails

AI Credits are new, they don’t get your capacity discount, and they aren’t included in the warehouse metrics teams typically watch. An agent or AI function that gets connected to a frequently running pipeline can scale its usage quickly. As such, it’s important to treat AI workloads like any other new cost center: set a budget, assign an owner, and review the usage regularly.

How to See What’s Driving Your Snowflake Bill

You can’t manage what you can’t see, and Snowflake gives you good raw material. In Snowsight, the Cost Management page breaks spend down by warehouse, service type, and day. For deeper analysis, these views are the ones we reach for first:

Two built-in controls help, with limits worth knowing. Resource monitors can notify or suspend warehouses at a credit threshold, but they only govern warehouses. Budgets cover supported serverless services too, but they send forecasts and notifications rather than capping spend.

Lowering Your Snowflake Costs

Understanding the pricing model is step one. Acting on it is step two: rightsizing warehouses, tuning auto-suspend, reviewing serverless features, and planning capacity against real usage. We cover each of those in depth in our guide to Snowflake cost optimization.

Final Thoughts on Snowflake Pricing

Although Snowflake pricing can be complex, it’s possible to understand the factors driving your compute consumption. The real challenge is keeping compute consumption efficient as workloads change.

Manual tuning works, but it’s periodic, and Snowflake workloads shift hour to hour. With AI-powered optimization and workload intelligence, you can identify high-impact savings and adjust warehouses automatically, around the clock.

Learn more about how Keebo helps eliminate wasted Snowflake spending and improve engineering efficiency.

Snowflake Pricing FAQs

How Much Does Snowflake Cost per Credit?

On-demand platform credits cost $2.00 to $3.10 on Standard, $3.00 to $4.65 on Enterprise, $4.00 to $6.20 on Business Critical, and $6.00 to $9.30 on VPS. Your cloud provider and region set where you fall in each range, and a capacity contract lowers the rate. AI Credits are priced separately at $2.00 to $2.20.

Is Snowflake Free to Use?

Snowflake offers a free trial with $400 of credits, so you can test the platform before committing. After that, you pay for what you use. The lowest rate is $2.00 per credit on Standard Edition in US regions, plus a monthly storage fee per terabyte of compressed data.

Is Snowflake Very Expensive?

Snowflake is not necessarily expensive. Because you pay for usage, your bill reflects how efficiently your workloads run, and costs can climb when warehouses are oversized or left running. That’s why Snowflake cost optimization is a priority at most organizations that use it heavily.

Is Snowflake Cost Effective?

Snowflake can be very cost effective when it’s actively managed. Teams with good visibility into usage, sensible warehouse settings, and a habit of acting on issues quickly tend to get strong value from it. Without that management, it’s easy to exceed your budget.

Do Idle Snowflake Warehouses Cost Money?

A suspended warehouse costs nothing. A running warehouse bills for every second it’s on, even if it isn’t processing queries. That’s why auto-suspend matters: it ensures a warehouse stops billing shortly after its last query.

Which Snowflake Edition Should I Choose?

Start with the features you need. Standard covers core warehousing. Enterprise adds multi-cluster warehouses, longer Time Travel, and advanced governance. Business Critical adds the security and compliance features regulated industries require. VPS is for organizations that need a fully isolated environment.

Should I Choose On-Demand or Capacity Pricing?

On-Demand suits new or unpredictable workloads, since you pay only for what you use. Capacity suits steady, forecastable workloads with annual spend above the $25,000 minimum, since it lowers your per-credit rate. Many teams start on On-Demand, then move to Capacity once they have several months of usage data to forecast from.

Do Unused Snowflake Capacity Credits Roll Over?

Sometimes. Snowflake says customers may be able to carry unused capacity into a future period, generally when they purchase additional capacity at renewal. The terms aren’t universal, so check your contract.

What’s Costing Me the Most in Snowflake?

For most accounts, it’s warehouse compute. Query WAREHOUSE_METERING_HISTORY to see which warehouses consume the most credits, then QUERY_HISTORY to find the queries driving them. Check METERING_HISTORY for serverless features you may have forgotten are running.

How Does Snowflake Billing Work?

On-Demand accounts are billed monthly in arrears for the previous month’s usage. Capacity accounts draw usage down against their prepaid commitment. For more detail, see our guide on how often Snowflake bills are generated.

How Does Databricks Pricing Compare to Snowflake?

Databricks charges for compute in Databricks Units (DBUs), with rates that vary by workload type. Storage, networking, and, in some configurations, your own cloud provider’s compute add separate costs. Snowflake charges credits for warehouse size and running time, plus storage on compressed data. Because DBUs and credits don’t represent equivalent compute, compare full workload bills rather than unit prices.

How Does Amazon Redshift Pricing Compare to Snowflake?

Amazon Redshift offers provisioned clusters billed per node-hour and a Serverless option billed per RPU-hour, with managed storage billed separately. Provisioned capacity can be reserved at a discount. Snowflake bills running warehouses in credits and charges separately for storage. Both models can suit steady or intermittent workloads; the cheaper option depends on your queries and how you configure each platform.

How Does BigQuery Pricing Compare to Snowflake?

Google BigQuery’s on-demand model charges $6.25 per TiB scanned, after the first 1 TiB each month. Its capacity model charges per slot-hour, starting at $0.04 on Standard edition. Storage bills separately. Snowflake charges for warehouse runtime rather than bytes scanned, so costs depend on warehouse size and how long it runs.

How Does Azure Synapse Pricing Compare to Snowflake?

Azure Synapse has two SQL pricing models: serverless SQL pools charge by data processed, and dedicated SQL pools charge for provisioned compute, with storage billed separately. Snowflake charges credits for warehouse runtime and separately for compressed storage. For lake queries, compare Synapse serverless scan charges against Snowflake warehouse time; for sustained warehouse workloads, compare dedicated-pool hours against Snowflake credits.